In the United States, diabetes mellitus is the leading cause of end-stage kidney failure, blindness in adults under age 65, and is the second leading cause (after trauma) of limb amputation. Poorly-controlled glucose levels are the major cause of these complications, but tight glycemic control is difficult to safely achieve using present technology. While a continuously-functioning glucose sensor would assist in safely achieving tight glucose control, such devices are typified by instability and loss of output over time due to the formation of foreign-body scar tissue, which eventually surrounds the sensors. Miniaturized sensors could be implanted under the skin with minimal discomfort. It is possible that the surrounding scar capsule could be made much more "friendly" to a glucose sensor by the slow release of growth factor compounds from the sensor surface. The hypothesis is that such compounds would reduce the scar fibrosis and generate many blood vessels in the capsule. These blood vessels are important to the long-term function of the sensor given its need for continuous glucose and oxygen delivery. A study will initially be performed in order to ascertain the optimal dosage of the growth factor. Then the growth factor (or saline control) will be slowly released over one month from miniaturized pumps into the tissue, which directly contacts the sensor. The thickness of the capsule and the formation of new blood vessels will measured by standard histologic staining techniques and by endothelial Factor Vifi immunohistochemistry. Another major problem of glucose sensors is short-term drift. It is frequently observed but poorly understood. It now appears that it may be possible to obtain stable readings from simultaneous recordings of multiple sensor electrodes. The technique which will be used to separate the accurate electrode signals from the inaccurate (outlying) signals is from the field of median statistics and is called the ZMAD method of Rousseeauw. The ZMAD data processing will be performed prospectively and continuously. In addition, the sensors and their transmitters will be miniaturized with the help of a biotelemetry company, MiniMitter. The body can be hostile to compounds which coat implanted devices. We will compare two promising polyurethanes as sensor coats: a carbonate based- vs. ether-based polymer. We will compare their long term function in studies using rabbits, which will be also be used to ascertain the effect of the growth factors and the real-time data processing.